The Oracle Singapore Needs But Won't Build
The Oracle Singapore Needs But Won’t Build
They merged WDA and SSG into a new super-bureau. Another committee. Another flowchart. Another layer of middle management that will spend six months aligning stakeholders before producing a PDF.
Meanwhile, the technology to give every Singaporean a personalized career optimizer—trained on local labor data, tuned to local industries, accessible via Singpass—is sitting on the table, production-ready, improving every second.
Let me be direct: instead of merging two statutory boards, they should have forked GLM 5.2, fine-tuned it on Singapore’s labor market, hooked it into Singpass, and let every citizen walk up to a centralized Oracle and ask “What should I do next?”
The Stack Is Ready
GLM 5.2 just dropped: 753B parameters, Mixture-of-Experts, IndexShare optimization, and a 1-million-token context window. That’s enough capacity to ingest every SkillsFuture course catalog, every industry manpower report, every JOBSTREET posting, and every MTI sector brief—and still have room for a citizen’s full conversation history.
AI Singapore already works in this sandbox with SEA-LION v4.5, adapting global open bases into regionally aware systems. The institutional muscle exists. The infrastructure patterns exist. The open-weight model exists.
The technical path is not speculative:
| Layer | What | How |
|---|---|---|
| Model | Fork GLM 5.2, MoE | Fine-tune on SG labor corpus |
| Identity | Singpass OAuth | Every citizen gets one session |
| Data | SkillsFuture, IRAS, MOM, MTI | Real-time feeds, not static PDFs |
| Inference | Self-hosted H100/H200 cluster | IMDA’s National LLM Programme budget |
| Surface | Web + WhatsApp + Telegram | Zero-friction access for all ages |
Three Reasons They Won’t Do It
1. The Stochastic Execution Problem
Governments run on deterministic logic. A policy is safe because it follows an audited flowchart: If X, apply subsidy Y.
Routing an entire population’s career transitions through a 753B MoE architecture means introducing non-deterministic execution at national scale. If the Oracle tells two identical engineers different things because of an emergent token alignment quirk, the state cannot defend that outcome in Parliament.
They would rather deploy a slower human committee—because a human committee provides a paper trail and a clear node of legal liability. Determinism isn’t a technical constraint for governments; it’s a psychological one.
2. The Compute Sovereignty Tax
Self-hosting a model of this footprint is expensive. An FP8 checkpoint of GLM 5.2 consumes roughly 860GB of VRAM, requiring dense H100 or H200 clusters just to sustain reasonable inference throughput for millions of concurrent users.
The civil service treats compute as a finite research commodity (rationed through IMDA’s programmes), not as an open utility. They would structurally prefer to subsidize private SaaS subscriptions—outsourcing both the cost and the liability.
3. The Truth Problem
An unvarnished Oracle optimized for pure efficiency would inevitably output truths that break the national narrative.
If the model analyzes real-time production data and concludes that a specific local industry is mathematically terminal, its advice to a 40-year-old worker logging in via Singpass might be:
“Stop spending your SkillsFuture credits on these courses. Your job category will be automated out of existence in 18 months. Maximize capital preservation immediately.”
The state cannot allow an automated system to undermine its core psychological framework of “continuous upskilling and career resilience.” The new SSG-WDA bureaucracy exists precisely to filter market realities through a lens of social stability—not to surface them raw.
The Actual Reason
The fundamental truth of “Singapore Inc.” is that the bureaucracy itself is a job-absorption machine. Building an elegant, centralized AI Oracle that automatically matches, guides, and orchestrates the workforce would render the entire middle management of the statutory boards structurally obsolete.
They didn’t merge WSG and SSG because it was the best engineering solution. They merged them because a centralized bureau is the only way a legacy government knows how to maintain the illusion of control.
But the tech is here. It keeps improving every second. And every year they spend aligning stakeholders is a year the Oracle could have been learning.
This post was distilled from a conversation with Gemini about the architectural gap between what Singapore’s labor machinery could be and what it is. The GLM 5.2 reference is current as of June 2026.